run_evaluate.py

# -*- coding: utf-8 -*-
"""
@author: Gabriel Mittag, TU-Berlin

This script evaluates a trained speech quality model on a given Dataset. 

If a 'csv_con' CSV file with per-condition results is provided, both the 
per-file CSV and per-condition CSV need a column with the name 'con' that 
contains the condition number. If 'csv_con' stays empty, only per-file results
are calculated.

"""
from nisqa.NISQA_model import nisqaModel

if __name__ == "__main__":

    args = {
            'mode': 'predict_csv',
            'pretrained_model': 'weights/nisqa.tar',
            'data_dir': r'C:\Users\Name\Downloads\NISQA_Corpus',
            'output_dir': r'C:\Users\Name\Downloads\NISQA_Corpus',
            'csv_file': 'NISQA_corpus_file.csv',
            'csv_con': 'NISQA_corpus_con.csv',
            'csv_deg': 'filepath_deg',
            'csv_mos_val': 'mos',
            'tr_num_workers': 6,
            'tr_bs_val': 40,
            'ms_channel': None,
            }
            
    nisqa = nisqaModel(args)
    nisqa.predict()
    nisqa.evaluate(
        mapping='first_order', 
        do_print=True,
        do_plot=True
        )